Re: Excess Phase........
This is exactly what I was trying to get at when saying (reducing phase wrap)
and when I mentioned other stuff relating to, do not alter the mic distance or software mic distance when measuring one to the next.
Except you managed to write it far more eloquently and clearly.
I must admit, when I first started measuring things it was like the blind leading the blind. I had not read anything as the 'hows' and 'whys' of doing certain things. And anything I head read wasnt quite clear enough.
I didn't realise (foolishly I admit, I was only 18 at the time) that you needed, as Dick so elegantly puts it,
That probably says it all when I say I went about it in a haphazard fashion with no relation between the measured drivers
It was no surprise that nothing made sense.
After that it all worked.
What type of consistency problems is it you are talking about? You measure the drivers one year and it works in a certain way. You measure them another year, and what you had before doesn't match up with what you have now?
Is it that you are trying to get perfect results every time you measure the drivers? But for some reasons 1 in 10 sets of measurements don't work? Id just dismiss that as some error somewhere and not give it a seconds thought.
dickmorgan22 said:
This is exactly what I was trying to get at when saying (reducing phase wrap)
and when I mentioned other stuff relating to, do not alter the mic distance or software mic distance when measuring one to the next.
Except you managed to write it far more eloquently and clearly.
I must admit, when I first started measuring things it was like the blind leading the blind. I had not read anything as the 'hows' and 'whys' of doing certain things. And anything I head read wasnt quite clear enough.
I didn't realise (foolishly I admit, I was only 18 at the time) that you needed, as Dick so elegantly puts it,
That probably says it all when I say I went about it in a haphazard fashion with no relation between the measured drivers
It was no surprise that nothing made sense.After that it all worked.
What type of consistency problems is it you are talking about? You measure the drivers one year and it works in a certain way. You measure them another year, and what you had before doesn't match up with what you have now?
Is it that you are trying to get perfect results every time you measure the drivers? But for some reasons 1 in 10 sets of measurements don't work? Id just dismiss that as some error somewhere and not give it a seconds thought.
Didn't have much time this weekend to measure but I did do the tweeter and mid to test the combined response.
This pic shows a deep null in the tweeter + mid response that doesn't match the combined.
Interesting thing is, I did the combined on the tweeter axis and this graph is with the mid set to 0,0,0 as well as the tweeter. If I change the DY position of the mid to where it actually is (130mm down), this graph gets way out of shape and doesn't shift the null at all.
Also, when adjusting the DY coords (for vertical driver positioning) switching between -130mm and +130mm for the mid position makes no difference. I assume to place a driver below that of the reference driver (0,0,0) I should be using -DY.
This pic shows a deep null in the tweeter + mid response that doesn't match the combined.
Interesting thing is, I did the combined on the tweeter axis and this graph is with the mid set to 0,0,0 as well as the tweeter. If I change the DY position of the mid to where it actually is (130mm down), this graph gets way out of shape and doesn't shift the null at all.
Also, when adjusting the DY coords (for vertical driver positioning) switching between -130mm and +130mm for the mid position makes no difference. I assume to place a driver below that of the reference driver (0,0,0) I should be using -DY.
Attachments
+/- 130 mm is a HUGE variation. Acoustic centers between drivers are typically no more than 15 mm different. You go through many phase rotations in a 260 mm span at 4 KHz, it is just pretty random.
I do not use LSP cad, so I cannot advise you on the tool. In general though, put your composite shot on the screen, along with the tweeter and midrange. Vary the Z-position (distance from the microphone) but NOT the X and Y since these were fixed for all three shots. Your computer does not care where they are, its fixed in the data taken. We are (I think??) trying to extract the acoustic center information so that your sims will work. This means figuring out what launch jitter your computer added, rather than changing anything else. The only thing that changed was the Z since your computer varied it for you with launch jitter.
Adjust the Z position until these match.....a couple of mm at a time. Its a tedious process. How far to vary is impossible to know, it depends on your computer's specific variations at the time the samples were taken. Continue until the summed tweeter and midrange matches the measured composite. THEN.....doe your simulations with the tweeter at 0,0,0, and the midrange (if thats what you varied) at 0,0,n where n is the number that you had to change Z distance to get the summation to work.
Dick
I do not use LSP cad, so I cannot advise you on the tool. In general though, put your composite shot on the screen, along with the tweeter and midrange. Vary the Z-position (distance from the microphone) but NOT the X and Y since these were fixed for all three shots. Your computer does not care where they are, its fixed in the data taken. We are (I think??) trying to extract the acoustic center information so that your sims will work. This means figuring out what launch jitter your computer added, rather than changing anything else. The only thing that changed was the Z since your computer varied it for you with launch jitter.
Adjust the Z position until these match.....a couple of mm at a time. Its a tedious process. How far to vary is impossible to know, it depends on your computer's specific variations at the time the samples were taken. Continue until the summed tweeter and midrange matches the measured composite. THEN.....doe your simulations with the tweeter at 0,0,0, and the midrange (if thats what you varied) at 0,0,n where n is the number that you had to change Z distance to get the summation to work.
Dick
Agreeing with everything Dick has said in his last post.
As an aside. The first graph you show, linking the tweeter and mid together is pretty much "perfect" in showing that the overall match you have is correct. The trend is obviously in the right direction and only varies by a slight amount.
The X and Y variables should not alter the response much at all, but the Z would make it shoot all over the place. What I mean here, is if you vary X or Y by 30mm, nothing much should change. But if you change Z by 30mm, lots will happen.
What you seem to be worried about is the depth of the notch at 4khz. My response is. Don't.
The notch is in the right place, and is also mirroring everything else going on around the notch almost perfectly.
If you have built a crossover and have observed the reverse null. Altering the crossover by a stupidly small amount can change your null from -50dB to -30dB. IE in the region of the tolerance of the cross over components. What I am saying is this difference in the depth of the notch will occur in areas out of your control.
As an aside. The first graph you show, linking the tweeter and mid together is pretty much "perfect" in showing that the overall match you have is correct. The trend is obviously in the right direction and only varies by a slight amount.
The X and Y variables should not alter the response much at all, but the Z would make it shoot all over the place. What I mean here, is if you vary X or Y by 30mm, nothing much should change. But if you change Z by 30mm, lots will happen.
What you seem to be worried about is the depth of the notch at 4khz. My response is. Don't.
The notch is in the right place, and is also mirroring everything else going on around the notch almost perfectly.
If you have built a crossover and have observed the reverse null. Altering the crossover by a stupidly small amount can change your null from -50dB to -30dB. IE in the region of the tolerance of the cross over components. What I am saying is this difference in the depth of the notch will occur in areas out of your control.
Thanks, Dick, Matt.
I'm prepared to look even more stupid: I moved the mic to the axis of the mid. I thought that is why I had to adjust the DY (up/down) position of the mid in software. I measured the combined response on the tweeter axis so I thought specifying the mid's vertical offset would allow the software to correctly match it to the combined response.
No?
I'm prepared to look even more stupid: I moved the mic to the axis of the mid. I thought that is why I had to adjust the DY (up/down) position of the mid in software. I measured the combined response on the tweeter axis so I thought specifying the mid's vertical offset would allow the software to correctly match it to the combined response.
No?
Actually, Matt, I've just gone and checked your suggested procedure on page 1 of this thread and you say to set the x and y position of the other drivers relative to the reference driver.
That's why I adjusted my mid by 130mm in the DY, because it's cone center is 130mm below the tweeter's, which was my reference axis for this measurement.
What am I misunderstanding?
That's why I adjusted my mid by 130mm in the DY, because it's cone center is 130mm below the tweeter's, which was my reference axis for this measurement.
What am I misunderstanding?
Do not adjust your set!
To get valid FR's usable for a crossover simulation, DO NOT move the microphone. You need a shot of the tweeter, a shot of the midrange, and a composite shot with both. You can take the woofer as well, but it makes it harder.
Set the mic at at least 1 meter distance, and do not movbe it when you take the tweeter, mid, and composite.
Your only goal here is to find the Z-axis offset that allows the mid and tweeter shots to add up to the composite shot. Leave that offset in place during all of your sims. I know its not really bthere, its a mathematical fuge that accounts for an error in your measument system. You can mod these number IF you want to change the baffle, but then I really recommend reshooting the drivers on the new baffle.
Dick
To get valid FR's usable for a crossover simulation, DO NOT move the microphone. You need a shot of the tweeter, a shot of the midrange, and a composite shot with both. You can take the woofer as well, but it makes it harder.
Set the mic at at least 1 meter distance, and do not movbe it when you take the tweeter, mid, and composite.
Your only goal here is to find the Z-axis offset that allows the mid and tweeter shots to add up to the composite shot. Leave that offset in place during all of your sims. I know its not really bthere, its a mathematical fuge that accounts for an error in your measument system. You can mod these number IF you want to change the baffle, but then I really recommend reshooting the drivers on the new baffle.
Dick
You Have it!
This process works, even if its a bit tedious. The good news is that you only have to do it once.
Now as to the woofers, for a typical 300 Hz crossover, I would just ignore it. In reality, you cannot measure the woofer response in this area indoors. You might get some decent readings outside, or in your neighborhood anechoic chamber.
You are using gated measurements right? Far field and near field are pretty useless for XO modeling. For the woofer, is a good idea to splice the near field and far field so that you have a clue what the LEVELS are. Other than that, I find low XO frequency designs are all about getting the level match first, and the slope right second. At 300 Hz you can't use a gated measurement, buts you can do in the nearfield. Phase is too hard to measure, so reverse it for 2nd order and keep it consistent for 4th order.
Dick
This process works, even if its a bit tedious. The good news is that you only have to do it once.
Now as to the woofers, for a typical 300 Hz crossover, I would just ignore it. In reality, you cannot measure the woofer response in this area indoors. You might get some decent readings outside, or in your neighborhood anechoic chamber.
You are using gated measurements right? Far field and near field are pretty useless for XO modeling. For the woofer, is a good idea to splice the near field and far field so that you have a clue what the LEVELS are. Other than that, I find low XO frequency designs are all about getting the level match first, and the slope right second. At 300 Hz you can't use a gated measurement, buts you can do in the nearfield. Phase is too hard to measure, so reverse it for 2nd order and keep it consistent for 4th order.
Dick
Dick, I am measuring outside.
And, I'm measuring at 1m, in fact I have things set up RIGHT NOW.
After I've got my mid & top end measured as stated, what next? Measure the farfield woofer response on the tweeter axis and adjust the nearfiled woofer response down to that level...?
Surely I need an accurate combined response INCLUDING the woofers?
And, I'm measuring at 1m, in fact I have things set up RIGHT NOW.
After I've got my mid & top end measured as stated, what next? Measure the farfield woofer response on the tweeter axis and adjust the nearfiled woofer response down to that level...?
Surely I need an accurate combined response INCLUDING the woofers?
Outside, the world is your oyster........
If you are outside, I assume you are doing a ground plane measurement. If you don't know what that is, I will walk you through it.
In this case, life is wonderful. I have to wait for a month without an "R" in it to do my outside measurements.....sniff. You do not need to do any nearfield measurements at all, and this is a good thing.
Do for the woofer as you have done for the tweeter and mid. Do not move anything, and measure all three drivers separately, then all at once. Get the mid/tweeter working first by adjusting bthe midrange Z distance. When the top end looks close, adjust the Z-axis offset for the woofer such that it sums with the mid and tweeter to match the composite picture you took of all three drivers active at once.
Outdoor measurements let you get very, very accurate results.
Dick
If you are outside, I assume you are doing a ground plane measurement. If you don't know what that is, I will walk you through it.
In this case, life is wonderful. I have to wait for a month without an "R" in it to do my outside measurements.....sniff. You do not need to do any nearfield measurements at all, and this is a good thing.
Do for the woofer as you have done for the tweeter and mid. Do not move anything, and measure all three drivers separately, then all at once. Get the mid/tweeter working first by adjusting bthe midrange Z distance. When the top end looks close, adjust the Z-axis offset for the woofer such that it sums with the mid and tweeter to match the composite picture you took of all three drivers active at once.
Outdoor measurements let you get very, very accurate results.
Dick
Dick, attached is a chart of tweeter + mid that I just did. Looks pretty good. I had to adjust the level of the mid down by about .5db and it all came into focus. The Z adjustment on the mid was only a couple of mm.
The stuff above 1k tracks nicely but below that it diverges. Is that just lack of resolution or can that be adjusted out as well?
Now, the woofers.
I have twin woofers so I just measure each separately on tweeter axis as with mid and then all four drivers together. Correct?
What about getting good resolution in the low end with nearfield on both the woofers and mid?
You didn't explain the ground plane measurement.
Thanks.
Mos
The stuff above 1k tracks nicely but below that it diverges. Is that just lack of resolution or can that be adjusted out as well?
Now, the woofers.
I have twin woofers so I just measure each separately on tweeter axis as with mid and then all four drivers together. Correct?
What about getting good resolution in the low end with nearfield on both the woofers and mid?
You didn't explain the ground plane measurement.
Thanks.
Mos
Attachments
Ground Plane Measurement
First off, when you set everything up, take all the measurements with NO adjustments whatsover to anything other than the amplifier driving the most efficient driver (usually the tweeter, to allow a level adjust) initially, then the other drivers in sequence with NO adjustments of anything mechanical, electrical, or cyber.
Ground plane measurements are the way to go. All measuring problems are related to in room reflections, or outside, to ground reflections. You can either do an ungated measurement, which is interesting but shows all the room modes, or you can do a gated measurement. Depending upon your "reflection free period" which is the time delay from the start to the end marker, you might get data down to 300 Hz. Typically its more like 500 Hz.
For those readers that are skeptical, if you have a 4 mSec "window" where your samples are considered valid, that is a 250 Hz wave with eaxctly ONE sample taken. Not very accurate at all.
To remove the gate problem you need to remove the reflections. Outside, the only place you have a problem is the ground. Putting the speaker on a 3 m step ladder and the mic on a 3 m extension will give you a nice wide window since the acoustic distance from driver to ground to mic is really quite long. This will get you down into the bass, but I don't like lining everything up 3 m off the ground.
So instead, do this. Lay a sheet of plywood on the ground. Anything acoustically reflective would work, with a sheet of glass being perfect. Just so its smooth. Put the speaker on a 1 m stand, although 2 m is a bit better. I use a step ladder and some bungee cords. Tilt it so that it is "pointed" at the sheet of plywood on the ground. The angle is not too critical, but aim for 45 deegrees. Lay your microphone right on the sheet of plywood. You are trying to minimize the distance between the microphone and the ground, hence the name "ground plane measurement".
The speaker on the ladder chirps down to the microphone on the ground, attached to an acousticly reflective surface. Now the reflection from the plywood arrives, effectively, with a zero delay to the incident wave.
This means you can ignore it.
This technique allows lower power (you get 3 dB gain from the plywood) and your stop marker can go out to infinity.
Start marker at distance to microphone, stop marker at infinity = Infinite bass resolution. And the phase data will be consistent and useful. And for the first time, ever, you will SEE the baffle step signature in all its glory.
Its a pain to set up, and you cannot do a good 3-way without it. I promise after going through the trouble, you will never do it any other way again. Good luck!
Dick
First off, when you set everything up, take all the measurements with NO adjustments whatsover to anything other than the amplifier driving the most efficient driver (usually the tweeter, to allow a level adjust) initially, then the other drivers in sequence with NO adjustments of anything mechanical, electrical, or cyber.
Ground plane measurements are the way to go. All measuring problems are related to in room reflections, or outside, to ground reflections. You can either do an ungated measurement, which is interesting but shows all the room modes, or you can do a gated measurement. Depending upon your "reflection free period" which is the time delay from the start to the end marker, you might get data down to 300 Hz. Typically its more like 500 Hz.
For those readers that are skeptical, if you have a 4 mSec "window" where your samples are considered valid, that is a 250 Hz wave with eaxctly ONE sample taken. Not very accurate at all.
To remove the gate problem you need to remove the reflections. Outside, the only place you have a problem is the ground. Putting the speaker on a 3 m step ladder and the mic on a 3 m extension will give you a nice wide window since the acoustic distance from driver to ground to mic is really quite long. This will get you down into the bass, but I don't like lining everything up 3 m off the ground.
So instead, do this. Lay a sheet of plywood on the ground. Anything acoustically reflective would work, with a sheet of glass being perfect. Just so its smooth. Put the speaker on a 1 m stand, although 2 m is a bit better. I use a step ladder and some bungee cords. Tilt it so that it is "pointed" at the sheet of plywood on the ground. The angle is not too critical, but aim for 45 deegrees. Lay your microphone right on the sheet of plywood. You are trying to minimize the distance between the microphone and the ground, hence the name "ground plane measurement".
The speaker on the ladder chirps down to the microphone on the ground, attached to an acousticly reflective surface. Now the reflection from the plywood arrives, effectively, with a zero delay to the incident wave.
This means you can ignore it.
This technique allows lower power (you get 3 dB gain from the plywood) and your stop marker can go out to infinity.
Start marker at distance to microphone, stop marker at infinity = Infinite bass resolution. And the phase data will be consistent and useful. And for the first time, ever, you will SEE the baffle step signature in all its glory.
Its a pain to set up, and you cannot do a good 3-way without it. I promise after going through the trouble, you will never do it any other way again. Good luck!
Dick
and by the way........
To answer your initial question, wire your woofer drivers the way you plan to use them, usually parallel, and measure them at the same time. Treat them as a single driver UNLESS you are doing a 3.5 way. This will matter if your Xover point is above 350 Hz. Below this point, treat 'em like a single driver.
Dick
To answer your initial question, wire your woofer drivers the way you plan to use them, usually parallel, and measure them at the same time. Treat them as a single driver UNLESS you are doing a 3.5 way. This will matter if your Xover point is above 350 Hz. Below this point, treat 'em like a single driver.
Dick
It just so happens the speaker IS sitting on a piece of plywood in the driveway right now. Problem is, the speaker weighs a ton, maybe 60-70kgs, very hard to lift onto anything--I have to take it outside on a furniture barrow
Can I do the ground plane with the speaker on the plywood? Even if I could get it up onto a table, it would only be about 700mm off the ground and then tilting it would be DANGEROUS, I'm sure.
I've seen pics in Stereophile where the mic is on the floor in front of the speaker. They must be doing a ground plane measurement there, and the speaker is just sitting normally on the floor about a metre from the mic.
Can I do the ground plane with the speaker on the plywood? Even if I could get it up onto a table, it would only be about 700mm off the ground and then tilting it would be DANGEROUS, I'm sure.
I've seen pics in Stereophile where the mic is on the floor in front of the speaker. They must be doing a ground plane measurement there, and the speaker is just sitting normally on the floor about a metre from the mic.
"It just so happens the speaker IS sitting on a piece of plywood in the driveway right now. Problem is, the speaker weighs a ton, maybe 60-70kgs, very hard to lift onto anything--I have to take it outside on a furniture barrow "
only 70 kilos?! Put your back into it man!
"Can I do the ground plane with the speaker on the plywood? Even if I could get it up onto a table, it would only be about 700mm off the ground and then tilting it would be DANGEROUS, I'm sure."
Yes, you can do this. My experience is that the top end gets fuzzy because some of the energy travels parallel to the ground plane when the angle is low, rather than bouncing cleanly off into the sky. Also, the microphone is not a point source......consider the implications for a bit. As a consequence, you will have some reflections that will show up, after the math happens, as high frequency noise superimposed on the real signal. Its way better than nothing.
Dick
"only 70 kilos?! Put your back into it man!
"Can I do the ground plane with the speaker on the plywood? Even if I could get it up onto a table, it would only be about 700mm off the ground and then tilting it would be DANGEROUS, I'm sure."
Yes, you can do this. My experience is that the top end gets fuzzy because some of the energy travels parallel to the ground plane when the angle is low, rather than bouncing cleanly off into the sky. Also, the microphone is not a point source......consider the implications for a bit. As a consequence, you will have some reflections that will show up, after the math happens, as high frequency noise superimposed on the real signal. Its way better than nothing.
Dick
Just gave it a go. Response attached. Response is for 2 x woofers+port
Mic was only 80cm from speaker. Speaker was not tilted. But, that looks pretty close to the manufacturer's (Vifa) specs for the M22 (above 1k, I mean), plus that looks like nice baffle step attenuation to me.
Baffle width is 300mm and I worked out some time ago that the BS -3db should be around 150hz.
So, how do I use this? Replace my farfield woofers response with the ground plane response? Problem I see there is that the combined response of all drivers includes the farfield response and not the ground plane response, so I won't be comparing apples with apples re the bottom end.
Mic was only 80cm from speaker. Speaker was not tilted. But, that looks pretty close to the manufacturer's (Vifa) specs for the M22 (above 1k, I mean), plus that looks like nice baffle step attenuation to me.
Baffle width is 300mm and I worked out some time ago that the BS -3db should be around 150hz.
So, how do I use this? Replace my farfield woofers response with the ground plane response? Problem I see there is that the combined response of all drivers includes the farfield response and not the ground plane response, so I won't be comparing apples with apples re the bottom end.
Attachments
Doesn't look quite bright yet
"Mic was only 80cm from speaker. Speaker was not tilted. But, that looks pretty close to the manufacturer's (Vifa) specs for the M22 (above 1k, I mean), plus that looks like nice baffle step attenuation to me."
You have to tilt the speaker to establish the design axis. It just works that way for ground plane measurements. Your baffle step signature does not look right.....its too small. This tells me that your mic was still in the near field (??) when you took the measurement. I'm not sure I understand what is going on preceisely, but its not right. The other thing bugging me is the appearance of your data. I've never used your package, but mine NEVER produces anything so pretty or so smooth.
"Baffle width is 300mm and I worked out some time ago that the BS -3db should be around 150hz."
Sure, but you should see an 8 dB step with a 6 dB overall level change. They always look like that, its just the frequency that changes. Yours is a half a step.
"So, how do I use this? Replace my farfield woofers response with the ground plane response? Problem I see there is that the combined response of all drivers includes the farfield response and not the ground plane response, so I won't be comparing apples with apples re the bottom end."
Back to the beginning. You set up the speaker, and you take four measurements without moving a thing or changing a setting. One for tweeter, midrange, woofer pair, and composite.
This is why you HAVE to tilt the speaker.....you can't do a ground plane measurement on the design axis otherwise. Seriously....give it some thought.....you MUST be a clever engineer to have got this far. Put a plank on two saw horses, stand the speaker on the plank facing outward along the plank axis. Staple a strap to the top of the speaker (thats why you measure before finishing) and tilt it facing downwards with the other end tied to the far end of the plank. The weight of the plank keeps the whole thing from tipping over.
Yeah, the box is heavy, and this requires some planning. You only need to set it up once, and you cannot really do a good three way any other way, unless you get very lucky. My $0.02 worth....
Dick
"Mic was only 80cm from speaker. Speaker was not tilted. But, that looks pretty close to the manufacturer's (Vifa) specs for the M22 (above 1k, I mean), plus that looks like nice baffle step attenuation to me."
You have to tilt the speaker to establish the design axis. It just works that way for ground plane measurements. Your baffle step signature does not look right.....its too small. This tells me that your mic was still in the near field (??) when you took the measurement. I'm not sure I understand what is going on preceisely, but its not right. The other thing bugging me is the appearance of your data. I've never used your package, but mine NEVER produces anything so pretty or so smooth.
"Baffle width is 300mm and I worked out some time ago that the BS -3db should be around 150hz."
Sure, but you should see an 8 dB step with a 6 dB overall level change. They always look like that, its just the frequency that changes. Yours is a half a step.
"So, how do I use this? Replace my farfield woofers response with the ground plane response? Problem I see there is that the combined response of all drivers includes the farfield response and not the ground plane response, so I won't be comparing apples with apples re the bottom end."
Back to the beginning. You set up the speaker, and you take four measurements without moving a thing or changing a setting. One for tweeter, midrange, woofer pair, and composite.
This is why you HAVE to tilt the speaker.....you can't do a ground plane measurement on the design axis otherwise. Seriously....give it some thought.....you MUST be a clever engineer to have got this far. Put a plank on two saw horses, stand the speaker on the plank facing outward along the plank axis. Staple a strap to the top of the speaker (thats why you measure before finishing) and tilt it facing downwards with the other end tied to the far end of the plank. The weight of the plank keeps the whole thing from tipping over.
Yeah, the box is heavy, and this requires some planning. You only need to set it up once, and you cannot really do a good three way any other way, unless you get very lucky. My $0.02 worth....
Dick
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.